Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: stephentoub <2642209+stephentoub@users.noreply.github.com>
509 lines
20 KiB
C#
509 lines
20 KiB
C#
using Microsoft.Extensions.DependencyInjection;
|
|
using ModelContextProtocol.Client;
|
|
using ModelContextProtocol.Protocol;
|
|
using ModelContextProtocol.Server;
|
|
using ModelContextProtocol.Tests.Utils;
|
|
using System.Text.Json;
|
|
|
|
namespace ModelContextProtocol.Tests.Server;
|
|
|
|
/// <summary>
|
|
/// Integration tests for task cancellation behavior, including TTL-based automatic
|
|
/// cancellation and explicit cancellation via tasks/cancel.
|
|
/// </summary>
|
|
public class TaskCancellationIntegrationTests : ClientServerTestBase
|
|
{
|
|
private readonly TaskCompletionSource<bool> _toolCancellationFired = new(TaskCreationOptions.RunContinuationsAsynchronously);
|
|
private readonly TaskCompletionSource<bool> _toolStarted = new(TaskCreationOptions.RunContinuationsAsynchronously);
|
|
|
|
public TaskCancellationIntegrationTests(ITestOutputHelper testOutputHelper)
|
|
: base(testOutputHelper)
|
|
{
|
|
}
|
|
|
|
protected override void ConfigureServices(ServiceCollection services, IMcpServerBuilder mcpServerBuilder)
|
|
{
|
|
// Add task store for server-side task support
|
|
var taskStore = new InMemoryMcpTaskStore();
|
|
services.AddSingleton<IMcpTaskStore>(taskStore);
|
|
|
|
services.Configure<McpServerOptions>(options =>
|
|
{
|
|
options.TaskStore = taskStore;
|
|
});
|
|
|
|
// Add a long-running tool that captures cancellation
|
|
mcpServerBuilder.WithTools([McpServerTool.Create(
|
|
async (CancellationToken ct) =>
|
|
{
|
|
_toolStarted.TrySetResult(true);
|
|
try
|
|
{
|
|
// Wait indefinitely until cancelled
|
|
await Task.Delay(Timeout.Infinite, ct);
|
|
return "completed";
|
|
}
|
|
catch (OperationCanceledException)
|
|
{
|
|
_toolCancellationFired.TrySetResult(true);
|
|
throw;
|
|
}
|
|
},
|
|
new McpServerToolCreateOptions
|
|
{
|
|
Name = "long-running-tool",
|
|
Description = "A tool that runs until cancelled"
|
|
})]);
|
|
}
|
|
|
|
private static IDictionary<string, JsonElement> EmptyArguments() => new Dictionary<string, JsonElement>();
|
|
|
|
[Fact]
|
|
public async Task TaskTool_CancellationToken_FiresWhenTtlExpires()
|
|
{
|
|
// Arrange
|
|
await using McpClient client = await CreateMcpClientForServer();
|
|
|
|
// Act - Call tool with short TTL (200ms)
|
|
var callResult = await client.CallToolAsync(
|
|
new CallToolRequestParams
|
|
{
|
|
Name = "long-running-tool",
|
|
Arguments = EmptyArguments(),
|
|
// Use a TTL long enough that thread pool scheduling delays on loaded CI machines
|
|
// don't cause the CTS to fire before the tool lambda begins executing.
|
|
Task = new McpTaskMetadata { TimeToLive = TimeSpan.FromSeconds(5) }
|
|
},
|
|
cancellationToken: TestContext.Current.CancellationToken);
|
|
|
|
// Verify task was created
|
|
Assert.NotNull(callResult.Task);
|
|
|
|
// Wait for the tool to start executing
|
|
await _toolStarted.Task.WaitAsync(TestConstants.DefaultTimeout, TestContext.Current.CancellationToken);
|
|
|
|
// Assert - Wait for the cancellation to fire (should happen when TTL expires)
|
|
var cancelled = await _toolCancellationFired.Task.WaitAsync(TestConstants.DefaultTimeout, TestContext.Current.CancellationToken);
|
|
Assert.True(cancelled, "Tool's CancellationToken should have been triggered when TTL expired");
|
|
|
|
// Note: TTL-based expiration does not explicitly set task status to Cancelled.
|
|
// Instead, expired tasks are considered "dead" and will be cleaned up by the task store.
|
|
// The task may still be in Working status or may throw "not found" if already cleaned up.
|
|
}
|
|
|
|
[Fact]
|
|
public async Task TaskTool_CancellationToken_FiresWhenExplicitlyCancelled()
|
|
{
|
|
// Arrange
|
|
await using McpClient client = await CreateMcpClientForServer();
|
|
|
|
// Start a long-running task with a long TTL
|
|
var callResult = await client.CallToolAsync(
|
|
new CallToolRequestParams
|
|
{
|
|
Name = "long-running-tool",
|
|
Arguments = EmptyArguments(),
|
|
Task = new McpTaskMetadata { TimeToLive = TimeSpan.FromMinutes(10) }
|
|
},
|
|
cancellationToken: TestContext.Current.CancellationToken);
|
|
|
|
Assert.NotNull(callResult.Task);
|
|
string taskId = callResult.Task.TaskId;
|
|
|
|
// Wait for the tool to start executing
|
|
await _toolStarted.Task.WaitAsync(TestConstants.DefaultTimeout, TestContext.Current.CancellationToken);
|
|
|
|
// Act - Explicitly cancel the task
|
|
var cancelledTask = await client.CancelTaskAsync(taskId, cancellationToken: TestContext.Current.CancellationToken);
|
|
|
|
// Assert - Wait for the cancellation to propagate to the tool
|
|
var cancelled = await _toolCancellationFired.Task.WaitAsync(TestConstants.DefaultTimeout, TestContext.Current.CancellationToken);
|
|
Assert.True(cancelled, "Tool's CancellationToken should have been triggered by explicit cancellation");
|
|
|
|
// Verify task status
|
|
Assert.Equal(McpTaskStatus.Cancelled, cancelledTask.Status);
|
|
}
|
|
|
|
[Fact]
|
|
public async Task TaskTool_CompletesSuccessfully_WhenNotCancelled()
|
|
{
|
|
// Arrange - Create a new test with a quick-completing tool
|
|
var quickToolCompleted = new TaskCompletionSource<string>(TaskCreationOptions.RunContinuationsAsynchronously);
|
|
|
|
var services = new ServiceCollection();
|
|
services.AddLogging();
|
|
var taskStore = new InMemoryMcpTaskStore();
|
|
services.AddSingleton<IMcpTaskStore>(taskStore);
|
|
|
|
var builder = services
|
|
.AddMcpServer()
|
|
.WithStreamServerTransport(
|
|
new System.IO.Pipelines.Pipe().Reader.AsStream(),
|
|
new System.IO.Pipelines.Pipe().Writer.AsStream());
|
|
|
|
builder.WithTools([McpServerTool.Create(
|
|
async (string input, CancellationToken ct) =>
|
|
{
|
|
await Task.Delay(50, ct); // Quick operation
|
|
var result = $"Result: {input}";
|
|
quickToolCompleted.TrySetResult(result);
|
|
return result;
|
|
},
|
|
new McpServerToolCreateOptions
|
|
{
|
|
Name = "quick-tool",
|
|
Description = "A tool that completes quickly"
|
|
})]);
|
|
|
|
services.Configure<McpServerOptions>(options =>
|
|
{
|
|
options.TaskStore = taskStore;
|
|
});
|
|
|
|
await using var client = await CreateMcpClientForServer();
|
|
|
|
// Act - Call tool with long TTL
|
|
var callResult = await client.CallToolAsync(
|
|
new CallToolRequestParams
|
|
{
|
|
Name = "long-running-tool", // Use the base class tool which will block
|
|
Arguments = EmptyArguments(),
|
|
Task = new McpTaskMetadata { TimeToLive = TimeSpan.FromMinutes(5) }
|
|
},
|
|
cancellationToken: TestContext.Current.CancellationToken);
|
|
|
|
Assert.NotNull(callResult.Task);
|
|
|
|
// Verify task is in working state initially
|
|
var task = await client.GetTaskAsync(callResult.Task.TaskId, cancellationToken: TestContext.Current.CancellationToken);
|
|
Assert.Equal(McpTaskStatus.Working, task.Status);
|
|
}
|
|
}
|
|
|
|
/// <summary>
|
|
/// Tests for task cancellation with multiple concurrent tasks.
|
|
/// </summary>
|
|
public class TaskCancellationConcurrencyTests : ClientServerTestBase
|
|
{
|
|
private readonly Dictionary<string, TaskCompletionSource<bool>> _toolCancellations = new();
|
|
private readonly Dictionary<string, TaskCompletionSource<bool>> _toolStarts = new();
|
|
private readonly object _lock = new();
|
|
|
|
public TaskCancellationConcurrencyTests(ITestOutputHelper testOutputHelper)
|
|
: base(testOutputHelper)
|
|
{
|
|
}
|
|
|
|
protected override void ConfigureServices(ServiceCollection services, IMcpServerBuilder mcpServerBuilder)
|
|
{
|
|
var taskStore = new InMemoryMcpTaskStore();
|
|
services.AddSingleton<IMcpTaskStore>(taskStore);
|
|
|
|
services.Configure<McpServerOptions>(options =>
|
|
{
|
|
options.TaskStore = taskStore;
|
|
});
|
|
|
|
// Tool that tracks cancellation per-invocation using a marker argument
|
|
mcpServerBuilder.WithTools([McpServerTool.Create(
|
|
async (string marker, CancellationToken ct) =>
|
|
{
|
|
TaskCompletionSource<bool> startTcs;
|
|
TaskCompletionSource<bool> cancelTcs;
|
|
|
|
lock (_lock)
|
|
{
|
|
if (!_toolStarts.TryGetValue(marker, out startTcs!))
|
|
{
|
|
startTcs = new TaskCompletionSource<bool>(TaskCreationOptions.RunContinuationsAsynchronously);
|
|
_toolStarts[marker] = startTcs;
|
|
}
|
|
if (!_toolCancellations.TryGetValue(marker, out cancelTcs!))
|
|
{
|
|
cancelTcs = new TaskCompletionSource<bool>(TaskCreationOptions.RunContinuationsAsynchronously);
|
|
_toolCancellations[marker] = cancelTcs;
|
|
}
|
|
}
|
|
|
|
startTcs.TrySetResult(true);
|
|
|
|
try
|
|
{
|
|
await Task.Delay(Timeout.Infinite, ct);
|
|
return $"completed-{marker}";
|
|
}
|
|
catch (OperationCanceledException)
|
|
{
|
|
cancelTcs.TrySetResult(true);
|
|
throw;
|
|
}
|
|
},
|
|
new McpServerToolCreateOptions
|
|
{
|
|
Name = "trackable-tool",
|
|
Description = "A tool that can be tracked by marker"
|
|
})]);
|
|
}
|
|
|
|
private void RegisterMarker(string marker)
|
|
{
|
|
lock (_lock)
|
|
{
|
|
_toolStarts[marker] = new TaskCompletionSource<bool>(TaskCreationOptions.RunContinuationsAsynchronously);
|
|
_toolCancellations[marker] = new TaskCompletionSource<bool>(TaskCreationOptions.RunContinuationsAsynchronously);
|
|
}
|
|
}
|
|
|
|
private Task WaitForStart(string marker, CancellationToken ct)
|
|
{
|
|
lock (_lock)
|
|
{
|
|
return _toolStarts[marker].Task.WaitAsync(TestConstants.DefaultTimeout, ct);
|
|
}
|
|
}
|
|
|
|
private Task<bool> WaitForCancellation(string marker, CancellationToken ct)
|
|
{
|
|
lock (_lock)
|
|
{
|
|
return _toolCancellations[marker].Task.WaitAsync(TestConstants.DefaultTimeout, ct);
|
|
}
|
|
}
|
|
|
|
private static IDictionary<string, JsonElement> CreateMarkerArgs(string marker) =>
|
|
new Dictionary<string, JsonElement>
|
|
{
|
|
["marker"] = JsonDocument.Parse($"\"{marker}\"").RootElement.Clone()
|
|
};
|
|
|
|
[Fact]
|
|
public async Task CancelTask_OnlyCancelsTargetTask_NotOtherTasks()
|
|
{
|
|
// Arrange
|
|
await using McpClient client = await CreateMcpClientForServer();
|
|
|
|
RegisterMarker("task1");
|
|
RegisterMarker("task2");
|
|
|
|
// Start two tasks
|
|
var result1 = await client.CallToolAsync(
|
|
new CallToolRequestParams
|
|
{
|
|
Name = "trackable-tool",
|
|
Arguments = CreateMarkerArgs("task1"),
|
|
Task = new McpTaskMetadata { TimeToLive = TimeSpan.FromMinutes(10) }
|
|
},
|
|
cancellationToken: TestContext.Current.CancellationToken);
|
|
|
|
var result2 = await client.CallToolAsync(
|
|
new CallToolRequestParams
|
|
{
|
|
Name = "trackable-tool",
|
|
Arguments = CreateMarkerArgs("task2"),
|
|
Task = new McpTaskMetadata { TimeToLive = TimeSpan.FromMinutes(10) }
|
|
},
|
|
cancellationToken: TestContext.Current.CancellationToken);
|
|
|
|
Assert.NotNull(result1.Task);
|
|
Assert.NotNull(result2.Task);
|
|
|
|
// Wait for both tools to start
|
|
await WaitForStart("task1", TestContext.Current.CancellationToken);
|
|
await WaitForStart("task2", TestContext.Current.CancellationToken);
|
|
|
|
// Act - Cancel only task1
|
|
await client.CancelTaskAsync(result1.Task.TaskId, cancellationToken: TestContext.Current.CancellationToken);
|
|
|
|
// Assert - task1 should be cancelled
|
|
var task1Cancelled = await WaitForCancellation("task1", TestContext.Current.CancellationToken);
|
|
Assert.True(task1Cancelled, "Task1 should have been cancelled");
|
|
|
|
// task2 should still be running (give it a moment to verify it wasn't cancelled)
|
|
var task2Status = await client.GetTaskAsync(result2.Task.TaskId, cancellationToken: TestContext.Current.CancellationToken);
|
|
Assert.Equal(McpTaskStatus.Working, task2Status.Status);
|
|
|
|
// Clean up - cancel task2
|
|
await client.CancelTaskAsync(result2.Task.TaskId, cancellationToken: TestContext.Current.CancellationToken);
|
|
}
|
|
|
|
[Fact]
|
|
public async Task MultipleTasks_WithDifferentTtls_CancelIndependently()
|
|
{
|
|
// Arrange
|
|
await using McpClient client = await CreateMcpClientForServer();
|
|
|
|
RegisterMarker("short-ttl");
|
|
RegisterMarker("long-ttl");
|
|
|
|
// Start task with short TTL. Use a TTL long enough that thread pool scheduling
|
|
// delays on loaded CI machines don't cause the CTS to fire before the tool
|
|
// lambda begins executing (CancelAfter starts counting at task creation, not
|
|
// when the tool's Task.Run is scheduled).
|
|
var shortTtlResult = await client.CallToolAsync(
|
|
new CallToolRequestParams
|
|
{
|
|
Name = "trackable-tool",
|
|
Arguments = CreateMarkerArgs("short-ttl"),
|
|
Task = new McpTaskMetadata { TimeToLive = TimeSpan.FromSeconds(5) }
|
|
},
|
|
cancellationToken: TestContext.Current.CancellationToken);
|
|
|
|
// Start task with long TTL
|
|
var longTtlResult = await client.CallToolAsync(
|
|
new CallToolRequestParams
|
|
{
|
|
Name = "trackable-tool",
|
|
Arguments = CreateMarkerArgs("long-ttl"),
|
|
Task = new McpTaskMetadata { TimeToLive = TimeSpan.FromMinutes(10) }
|
|
},
|
|
cancellationToken: TestContext.Current.CancellationToken);
|
|
|
|
Assert.NotNull(shortTtlResult.Task);
|
|
Assert.NotNull(longTtlResult.Task);
|
|
|
|
// Wait for both to start
|
|
await WaitForStart("short-ttl", TestContext.Current.CancellationToken);
|
|
await WaitForStart("long-ttl", TestContext.Current.CancellationToken);
|
|
|
|
// Assert - short TTL task should be cancelled automatically
|
|
var shortCancelled = await WaitForCancellation("short-ttl", TestContext.Current.CancellationToken);
|
|
Assert.True(shortCancelled, "Short TTL task should have been cancelled when TTL expired");
|
|
|
|
// Long TTL task should still be running
|
|
var longTtlStatus = await client.GetTaskAsync(longTtlResult.Task.TaskId, cancellationToken: TestContext.Current.CancellationToken);
|
|
Assert.Equal(McpTaskStatus.Working, longTtlStatus.Status);
|
|
|
|
// Clean up
|
|
await client.CancelTaskAsync(longTtlResult.Task.TaskId, cancellationToken: TestContext.Current.CancellationToken);
|
|
}
|
|
}
|
|
|
|
/// <summary>
|
|
/// Tests verifying that terminal task states (completed, failed, cancelled) cannot transition.
|
|
/// Per spec: "Tasks with a completed, failed, or cancelled status are in a terminal state
|
|
/// and MUST NOT transition to any other status"
|
|
/// </summary>
|
|
public class TerminalTaskStatusTransitionTests : ClientServerTestBase
|
|
{
|
|
public TerminalTaskStatusTransitionTests(ITestOutputHelper testOutputHelper)
|
|
: base(testOutputHelper)
|
|
{
|
|
}
|
|
|
|
protected override void ConfigureServices(ServiceCollection services, IMcpServerBuilder mcpServerBuilder)
|
|
{
|
|
var taskStore = new InMemoryMcpTaskStore();
|
|
services.AddSingleton<IMcpTaskStore>(taskStore);
|
|
|
|
services.Configure<McpServerOptions>(options =>
|
|
{
|
|
options.TaskStore = taskStore;
|
|
});
|
|
|
|
mcpServerBuilder.WithTools([
|
|
McpServerTool.Create(
|
|
async (CancellationToken ct) =>
|
|
{
|
|
await Task.Delay(10, ct);
|
|
return "quick result";
|
|
},
|
|
new McpServerToolCreateOptions
|
|
{
|
|
Name = "quick-tool",
|
|
Description = "A tool that completes quickly"
|
|
}),
|
|
McpServerTool.Create(
|
|
async (CancellationToken ct) =>
|
|
{
|
|
await Task.Delay(10, ct);
|
|
throw new InvalidOperationException("Intentional failure");
|
|
#pragma warning disable CS0162
|
|
return "never";
|
|
#pragma warning restore CS0162
|
|
},
|
|
new McpServerToolCreateOptions
|
|
{
|
|
Name = "failing-tool",
|
|
Description = "A tool that always fails"
|
|
})
|
|
]);
|
|
}
|
|
|
|
private static IDictionary<string, JsonElement> EmptyArguments() => new Dictionary<string, JsonElement>();
|
|
|
|
[Fact]
|
|
public async Task CompletedTask_CannotTransitionToOtherStatus()
|
|
{
|
|
// Arrange
|
|
await using McpClient client = await CreateMcpClientForServer();
|
|
|
|
var callResult = await client.CallToolAsync(
|
|
new CallToolRequestParams
|
|
{
|
|
Name = "quick-tool",
|
|
Arguments = EmptyArguments(),
|
|
Task = new McpTaskMetadata()
|
|
},
|
|
cancellationToken: TestContext.Current.CancellationToken);
|
|
|
|
Assert.NotNull(callResult.Task);
|
|
string taskId = callResult.Task.TaskId;
|
|
|
|
// Wait for completion
|
|
McpTask taskStatus;
|
|
do
|
|
{
|
|
await Task.Delay(50, TestContext.Current.CancellationToken);
|
|
taskStatus = await client.GetTaskAsync(taskId, cancellationToken: TestContext.Current.CancellationToken);
|
|
}
|
|
while (taskStatus.Status == McpTaskStatus.Working);
|
|
|
|
Assert.Equal(McpTaskStatus.Completed, taskStatus.Status);
|
|
|
|
// Act - Try to cancel a completed task (should be idempotent)
|
|
var cancelResult = await client.CancelTaskAsync(taskId, cancellationToken: TestContext.Current.CancellationToken);
|
|
|
|
// Assert - Status should still be completed (not cancelled)
|
|
Assert.Equal(McpTaskStatus.Completed, cancelResult.Status);
|
|
|
|
// Verify via get
|
|
var verifyStatus = await client.GetTaskAsync(taskId, cancellationToken: TestContext.Current.CancellationToken);
|
|
Assert.Equal(McpTaskStatus.Completed, verifyStatus.Status);
|
|
}
|
|
|
|
[Fact]
|
|
public async Task FailedTask_CannotTransitionToOtherStatus()
|
|
{
|
|
// Arrange
|
|
await using McpClient client = await CreateMcpClientForServer();
|
|
|
|
var callResult = await client.CallToolAsync(
|
|
new CallToolRequestParams
|
|
{
|
|
Name = "failing-tool",
|
|
Arguments = EmptyArguments(),
|
|
Task = new McpTaskMetadata()
|
|
},
|
|
cancellationToken: TestContext.Current.CancellationToken);
|
|
|
|
Assert.NotNull(callResult.Task);
|
|
string taskId = callResult.Task.TaskId;
|
|
|
|
// Wait for failure
|
|
McpTask taskStatus;
|
|
do
|
|
{
|
|
await Task.Delay(50, TestContext.Current.CancellationToken);
|
|
taskStatus = await client.GetTaskAsync(taskId, cancellationToken: TestContext.Current.CancellationToken);
|
|
}
|
|
while (taskStatus.Status == McpTaskStatus.Working);
|
|
|
|
Assert.Equal(McpTaskStatus.Failed, taskStatus.Status);
|
|
|
|
// Act - Try to cancel a failed task (should be idempotent)
|
|
var cancelResult = await client.CancelTaskAsync(taskId, cancellationToken: TestContext.Current.CancellationToken);
|
|
|
|
// Assert - Status should still be failed
|
|
Assert.Equal(McpTaskStatus.Failed, cancelResult.Status);
|
|
}
|
|
}
|